A couple of my users have recently reported problems with stale NFS file
handles - unfortunately the problem seems to be 'transient', in that
when I try and investigate the 'problem' - it's gone away ...
The work flow is basically a number of NFS clients render images to an
NFS server (which may a users' workstation) i.e. many clients writing
different files to an NFS mounted directory. Very occasionally one or
more render will fail and log an 'stale NFS file handle'. However
re-rendering the frame (which may or may not be from the same client)
usual works OK.
One user did have a problem whereby writing over a given number of files
to a directory failed with stale NFS file handle i.e. re-renders failed.
This problem was 'fixed' by removing the directory and its contents,
creating a new directory (on the same server) and rendering again.
All the servers/clients are running a mixture of XFS v1.1 (2.4.18
kernel) and v1.0.2 (2.4.7 kernel) - the problem appears to be
independent of client/server combinations.
Again, I don't know if this is an NFS or XFS problem - it's very rare
and I can't find anything out of the ordinary in various logs on the
clients or servers.
I know there are/have been NFS/XFS issues - could these show up as
transient stale NFS file handles?
Thanks
James Pearson
|